Navigation Visual Question Answering on NaVQA Medium (M) horizon
73.7Descriptive AccuracyReMEmbR
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ReMEmbRLLM=Gemini-3-Pro, Encoder=MixedBread2026.06 | 73.7 | 24.52 | 1.08 | |
| ReMEmbRLLM=Gemini-2.5-Flash, Encoder=MixedBread2026.06 | 60.5 | 28.89 | 2.01 | |
| RAVENLLM=Gemini-2.5-Pro, Encoder=QQMM-v2, Quantization=4-bit2026.06 | 60.5 | 12.29 | 0.69 | |
| VLM OnlyLLM=Gemini-2.5-Pro2026.06 | 60.5 | 12.35 | 0.98 | |
| VLM OnlyLLM=Gemini-2.5-Flash2026.06 | 50 | 12.48 | 1.19 | |
| RAVENLLM=Gemini-2.5-Flash, Encoder=QQMM-v2, Quantization=4-bit2026.06 | 47.4 | 16.2 | 0.67 | |
| RAVENLLM=Gemini-2.5-Flash, Encoder=Seed1.6-Embed2026.06 | 42.4 | 60.51 | 1.4 |